(a) Technical Field
The present invention relates to a system and method for recognizing a parking space line marking for a vehicle that automatically extracts and recognizes the parking space line markings from an image of a vehicular imaging device, such as a camera, video camera, vision system, etc.
(b) Background Art
In recent years, as user's interest in automatic parking has increased, research in this technology area has begun to play a major role in the business sector and academic sector as well as the research and development sector of the automotive industry. Such an automatic parking system is conventionally made up of three components such as target position designation process, a path planning process, and a path tracking process all of which are executed by a processor that is typically part of a controller installed in the vehicle. Conventional target position designation process may be classified into four categories which include a user-interface based process, an infrastructure based process, a free-space based process, and a parking space line marking based process. Most products produced at present use a user-interface based process or a free-space based process.
The user-interface based process is the most easily implemented and is useful for correction when automatic recognition has failed, but the user must manually operate the system every time and thus it is not a desirable system to most users. The free-space based process uses an ultrasonic sensor to provide a practical solution to the user during parallel parking. However, a the systems error rate during perpendicular parking is unreasonably high in comparison and as a result the targeted parking position is often incorrectly set obliquely depending on the height of the directly adjacent vehicles.
A parking space line marking recognition based process, however, has the following advantages as compared with other processes.
1) The parking space line marking recognition based process can minimize the operation degree of a driver which is a very significant problem in a user-interface based process. 2) Since the free-space based process sets a parking location depending on the height of a directly adjacent vehicle, a target location may be set obliquely. In the parking space line marking recognition process, a target may be set to be accurately arranged between on a parking space line marking lines. 3) Scanning laser radar, short range radar, and stereo camera based processes typically need to include additional sensors, however the parking space line marking recognition process may use an existing rear imaging device. 4) A process using an ultrasonic sensor has a high rate of error during perpendicular parking due to a sensing limit, but the parking space line marking recognition process can be effectively applied during perpendicular parking. 5) A monocular motion stereo based process requires a large amount of computing, but the parking space line marking recognition process may be implemented with a comparatively small amount of computing.
Yet still, the parking space line marking recognition based process can only be used when the parking space line markings are provided. However, most of the time in urban environments (where consumers are most likely to utilize this feature), parking spaces are adequately marked and therefore, this disadvantage is moot. The parking space line marking based process recognizes parking space line markings via only an image sensor as long as an infrastructure is not installed at the time of drawing the parking space line marking.
The parking space line marking based process may be classified into a semi-automatic process requiring a user input for information regarding a parking space and a full-automatic process which does not require user input. The semi-automatic process may be more preferably in terms of a velocity and recognition rate than the full-automatic process because the user inputs additional information related to a parking space and if an automatic recognition system fails, the semi-automatic process may be highly valuable even as a backup tool.
In some system, when the user inputs one point in a zone where the user wants to park the car through a touch screen, a method of recognizing the parking space by a marking line-segment with a directional intensity gradient based on the information is provided. However, this process is applied to only one kind of parking space line marking Therefore, the initial position of two cross points indicating an entrance of the parking zone which the user has requested is input in the touch screen in order to complement it, and a process of recognizing the parking space by template-matching a skeleton of a parking space line marking area acquired through binarization is provided.
The full-automatic process has been developed conventionally in a variety of manners. In some systems, there is provided a process of recognizing the parking space by extracting pixels included in the parking space line marking through a color segmentation process based on a restricted Coulomb energy neural network by using reasoning that the parking space line marking has a predetermined color.
In addition, after 3D reconstruction for a vertical edge in the rear of the vehicle is performed by using a binocular stereo camera, the parking space line marking is identified by using only the pixels that satisfy homography with a ground surface.
Meanwhile, there is provided additionally a process of recognizing the parking space by filtering a Hough space of a bird's-eye view edge image under the assumption that the parking space line marking is configured with two straight lines having a fixed width, which are parallel to each other and a straight line perpendicular thereto. Also there is provided a process of removing an out lying marking and detecting a straight line through an improved random sample consensus algorithm after continuously calculating a virtual target position based on a vehicle position and a deflection angle just before the vehicle starts moving rearward for parking. However, the existing full-automatic processes assume that there is only one type of parking space and thus is limited in its adaptability to unconventional spaces.
The description provided above as a related art is just for helping understanding the background of the present invention and should not be construed as being included in the related art known by those skilled in the art.